Adaptive control system for integrated photovoltaic-storage-charging station considering source-load uncertainty

By constructing a global operating condition state space and an adaptive control operator, the problem of high-frequency power surges in the integrated photovoltaic-storage-charging station under source-load uncertainty was solved, achieving efficient system operation and equipment lifespan protection in extreme environments.

CN123121771APending Publication Date: 2026-11-17GUANGXI TECHCAL COLLEGE OF MACHINERY & ELECTRICITY +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611594899.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-15
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

The existing integrated photovoltaic, energy storage and charging station control system cannot effectively mitigate high-frequency power surges when facing source load uncertainties, causing the energy storage components to bear excessive transient physical stress, which limits the adaptive adjustment efficiency of the control system in extreme environments.

Method used

By constructing a global operating state space, the electrophysical characteristic signals of the photovoltaic-storage-charging integrated station are extracted, and the probability interval prediction of source-load power is performed. Combining economic operating costs and energy storage losses, the power allocation command is optimized, and the control gain is dynamically adjusted using an adaptive control operator to generate converter drive signals, thereby achieving flexible and smooth processing of energy storage.

Benefits of technology

It achieves a balance between operating efficiency and equipment lifespan under extreme conditions, ensuring the physical safety and service life of energy storage in complex and random environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123121771A_ABST
    Figure CN123121771A_ABST
Patent Text Reader

Abstract

This invention discloses an adaptive control system for an integrated photovoltaic-storage-charging station that considers source-load uncertainty, relating to the field of power supply system technology. The system includes: a prediction module that predicts the probabilistic interval of source-load power based on the global operating state space, obtaining a predicted source-load distribution sequence and identifying the expected power deviation due to source-load randomness; an optimization module that combines the predicted source-load distribution sequence and the expected power deviation, optimizing with the goal of minimizing economic operating costs and energy storage losses, outputting power allocation commands and an adaptive control operator; and an execution module that dynamically adjusts the control gain according to the adaptive control operator to smooth the power allocation commands and reduce the transient physical stress of uncertainty fluctuations on energy storage, generating a converter drive signal. This achieves the lowest possible operating cost throughout the entire lifecycle while ensuring the physical safety and service life of energy storage in complex and random environments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power supply system technology, and in particular to an adaptive control system for an integrated photovoltaic-storage-charging station that takes into account the uncertainty of source and load. Background Technology

[0002] With the continuous increase in the proportion of distributed energy access, integrated photovoltaic-storage-charging stations, as a key form of microgrids, have become an important means to alleviate grid load pressure and improve renewable energy consumption. Existing control technologies mostly adopt a hierarchical energy management framework based on deterministic models. This framework collects static operating parameters such as photovoltaic power, energy storage state of charge, and charging load, and uses linear programming or model predictive control algorithms to generate power dispatch commands. These technologies typically rely on preset fixed system parameters and point-predicted values, establishing energy balance equations to achieve power allocation and grid-connected power regulation among units within the station. This has played a significant role in smoothing load fluctuations and improving the economic operation of the system.

[0003] However, under actual operating conditions, due to the sudden and drastic changes in the meteorological environment and the strong randomness of charging behavior, the system faces highly complex dual uncertainties of source and load. Existing control schemes often treat the energy storage system as an ideal power buffer unit, lacking the ability to extract the dynamic physical characteristics during power conversion online. This results in the control strategy being unable to dynamically correct itself based on the real-time electrochemical state evolution of the energy storage device. When the source and load power deviation exceeds the predicted range, the fixed control gain is insufficient to effectively mitigate high-frequency power surges. This not only easily leads to an increase in the residual between the dispatch command and the actual response but also subjectes the energy storage elements to excessive transient physical stress, thereby limiting the adaptive adjustment performance of the control system under extreme uncertainty environments. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an adaptive control system for an integrated photovoltaic-storage-charging station that considers source-load uncertainty to solve the problem of balancing source-load uncertainty and dynamic physical response of energy storage.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides an adaptive control system for an integrated photovoltaic-storage-charging station that considers source-load uncertainty, comprising: The state construction module acquires real-time operating data of the integrated photovoltaic-storage-charging station, extracts electrophysical characteristic signals during the power conversion process to identify the dynamic equivalent parameters of energy storage online, and constructs a global operating condition state space. The prediction module performs probability interval prediction of source load power based on the global operating condition state space, obtains the source load prediction distribution sequence, and identifies the expected value of power deviation caused by the randomness of source load. The optimization module combines the source-load predicted distribution sequence with the expected power deviation value to optimize the economic operating cost and energy storage loss, and outputs power allocation instructions and adaptive control operators. The execution module dynamically adjusts the control gain according to the adaptive control operator to smooth the power distribution command and reduce the transient physical stress of uncertainty fluctuations on energy storage, and generates the converter drive signal. The self-evolution module monitors the feedback state after the converter drive signal is applied and calculates the deviation residual, dynamically correcting the correlation parameters between the global operating condition state space and the optimization objective.

[0007] Preferably, the method for constructing the global operating condition state space includes: The electrical parameters of the energy exchange nodes in the integrated photovoltaic, energy storage and charging station are collected to obtain real-time operating data, and the data is aligned with the ambient temperature data on a time scale. Modal decomposition is performed on real-time operating data to separate the high-frequency components generated by switching actions and the low-frequency components of power fluctuations, and the high-frequency components are identified as electrophysical characteristic signals. Using electrophysical characteristic signals as excitation observations, and leveraging their transient response characteristics, a mapping relationship between current variation rate and voltage drop is established. The physical impedance parameters inside the energy storage are determined through online iterative convergence to identify dynamic equivalent parameters. Furthermore, multi-dimensional vector fusion is performed with real-time operating data to construct a global operating condition state space that reflects the energy exchange state and physical degradation trend.

[0008] Preferably, the method for predicting the probability interval of source-load power includes: The low-frequency components are serialized and correlated with real-time operating data on the time axis, and dynamic equivalent parameters are fused to extract the spatiotemporal features of source load evolution. Based on the historical evolution trajectory of the source-load evolution spatiotemporal characteristics, the expected trajectory of future power is extrapolated, and the prediction error sequence is determined by comparing with the historical measured power, thus obtaining the probability interval of the source-load power. Within the prediction bandwidth formed by the probability interval, based on the statistical distribution characteristics of the prediction error sequence, multiple random power trajectories fluctuating around the expected future power trajectory are generated as typical power evolution sequences and arranged according to probability weights to obtain the source-load prediction distribution sequence.

[0009] Preferably, the method for identifying the expected power deviation due to source-load randomness includes: The typical power evolution sequence is synchronized with the expected future power trajectory, and the real-time offset vector of the typical power evolution sequence relative to the expected future power trajectory is extracted. The real-time offset vector is mapped and associated with the probability weights corresponding to typical power evolution sequences to determine the contribution of each typical power evolution sequence to power uncertainty. Based on the degree of contribution, the real-time offset vectors of each typical power evolution sequence at the same prediction time are weighted and integrated to identify the expected power deviation caused by the randomness of source and load.

[0010] Preferably, the output power allocation command is provided in the following manner: Based on each typical power evolution sequence and its probability weight, and combined with the station's operating loss parameters, an economic operating cost function for operational efficiency is constructed. By utilizing the expected value of power deviation and dynamic equivalent parameters, the physical stress damage of random source load fluctuations on the internal physical impedance of energy storage is quantified, and an energy storage loss cost function is constructed. The economic operating cost function and the energy storage loss cost function are weighted and integrated to form a global objective function. Then, an optimization decision space is established by combining real-time operating data to find the optimal power flow scheme for each branch and output power allocation instructions.

[0011] Preferably, the method for obtaining the adaptive control operator includes: By using dynamic equivalent parameters and power allocation commands, the baseline control strength required to suppress power fluctuations is determined; the slope of the expected power deviation changes over time is analyzed to extract dynamic fluctuation characteristics that characterize the severity of random disturbances in the source load. The dynamic fluctuation characteristics are compared with the baseline control strength. When the dynamic fluctuation characteristics exceed the preset safe fluctuation threshold, the correction deviation of the control gain is determined. The correction deviation is then added to the baseline control strength to output the adaptive control operator.

[0012] Preferably, the method of dynamically adjusting the control gain according to the adaptive control operator includes: Using the power distribution command as a reference trajectory, the vector alignment with the real-time operating data is performed on a time scale to extract the initial error state of the converter's current regulation performance and the degree of environmental disturbance. The initial error state is input into the adaptive control operator and compared with the hardware control boundary to determine the gain adjustment coefficient for adjusting the sensitivity of the underlying loop. By using the gain adjustment coefficient to update the control parameters of the converter's internal loop in real time, the control gain can be dynamically adjusted according to the adaptive control operator.

[0013] Preferably, the method for generating the converter drive signal includes: The control loop bandwidth of the converter is adjusted by using the dynamically adjusted control gain, and the power distribution command is smoothed by filtering to obtain a smooth power reference. By comparing real-time operating data with a smoothed power reference, a compensation current vector is generated to offset transient current surges. This vector is then superimposed on the target control voltage converted from the smoothed power reference to form a comprehensive control voltage command. The comprehensive control voltage command is compared with the preset high-frequency carrier signal in real time, and the polarity of the output level is triggered to generate the converter drive signal based on the amplitude relationship between the comprehensive control voltage command and the preset high-frequency carrier signal.

[0014] Preferably, the method for calculating the deviation residual includes: Monitor the feedback state after the converter drive signal is applied; and perform multi-dimensional vector alignment with the expected response state composed of power allocation command and future power expectation trajectory to determine the multi-dimensional error vector; By using dynamic equivalent parameters, the steady-state component caused by inherent energy storage losses is separated from the initial error vector, and the dynamic deviation characteristics caused by random disturbances of source and load are extracted to obtain the deviation residual.

[0015] Preferably, the correlation parameters between the corrected global operating condition state space and the optimization objective include: The deviation residual is mapped to the global operating condition state space to determine the correction vector; and then superimposed onto the global operating condition state space to update the weights of the electrophysical characteristic signals in the global operating condition state space. Based on the deviation residuals, the weight allocation of the global objective function is adjusted according to the proportion of economic operating costs and energy storage losses to determine the sensitivity coefficient; combined with the calibrated global operating state space, the input benchmark and target constraints of the optimization action are updated synchronously to achieve dynamic correction of the associated parameters.

[0016] The beneficial effects of this invention are as follows: By extracting high-frequency electrophysical characteristic signals to identify the dynamic equivalent parameters of energy storage online and constructing a global operating condition state space, and by quantifying the stress damage caused by random fluctuations to physical impedance to construct an energy storage loss cost function, a logical closed loop of "deep physical state perception" and "precise stress loss offsetting" is realized. Its function is to transform the fuzzy randomness of source and load into quantifiable physical stress indicators, and to use an adaptive control operator to drive the converter for flexible smoothing and transient current compensation. This is used to solve the conflict between operating efficiency and equipment lifespan in photovoltaic-storage-charging integrated stations under extreme operating conditions, enabling the control system to have "physical self-sensing" and "operational self-evolution" capabilities. While minimizing the operating cost throughout the entire life cycle, it ensures the physical safety and service life of energy storage in complex random environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the adaptive control system for the integrated photovoltaic-storage-charging station that considers source-load uncertainty in this invention.

[0019] Figure 2 This is a flowchart illustrating the process of obtaining the source load prediction distribution sequence in this invention.

[0020] Figure 3 This is a flowchart of the output adaptive control operator in this invention.

[0021] Figure 4 This is a flowchart of the self-evolution module in this invention. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figure 1 , Figure 2 , Figure 3 and Figure 4 An embodiment of the present invention provides an adaptive control system for an integrated photovoltaic-storage-charging station that considers source-load uncertainty, comprising the following steps: Methods for constructing the global operating condition state space include: Electrical parameters of energy exchange nodes within the integrated photovoltaic, energy storage, and charging station are collected to obtain real-time operational data, which is then aligned with ambient temperature data over time.

[0026] It should be noted that the three-phase voltage and three-phase current of the photovoltaic inverter access point, energy storage converter access point, and AC bus node are captured in real time by the electrical sensors of the energy exchange nodes in the integrated photovoltaic-storage-charging station to obtain real-time operation data; the ambient temperature in the station is obtained by using an external ambient temperature sensor, and the real-time operation data and ambient temperature data are timestamped and aligned based on the time clock generated by a high-precision crystal oscillator. Specifically, the timing clock generated by a high-precision crystal oscillator is exemplified by a frequency stability that achieves... The microsecond-level pulse sequence is generated by a temperature-controlled crystal oscillator of order magnitude. A unique absolute time tag is assigned to each frame of the sampled signal in the real-time operation data by the temperature-controlled crystal oscillator, and the low-frequency ambient temperature data collected by the ambient temperature sensor is mapped according to the same absolute time tag, so as to achieve millisecond-level synchronization error alignment between the real-time operation data and the ambient temperature data on a unified time axis.

[0027] Modal decomposition is performed on real-time operating data to separate the high-frequency components generated by switching actions and the low-frequency components of power fluctuations, and the high-frequency components are identified as electrophysical characteristic signals.

[0028] Specifically, by setting the number of decomposition layers, with an example value of 5 layers and a preset penalty factor of 2000, a constrained variational problem mapping relationship is constructed. In the frequency domain, an iterative optimization strategy is used to continuously update the center frequency and bandwidth of each modal component, so that the sum of the estimated bandwidths of each modal component is minimized, thereby performing multi-scale decomposition on the sampled current sequence in the real-time running data. Based on the center frequency distribution characteristics of each modal component after decomposition, signal components in the real-time operating data whose frequency is within the converter pulse width modulation switching frequency (example value 10kHz) and offset by 20% above and below are identified as high-frequency components. Signal components in the real-time operating data whose frequency is within the range of power grid fluctuations and load variations (example value 0.1Hz to 50Hz) are identified as low-frequency components. The amplitude and phase fluctuation characteristics of the high-frequency components are extracted and identified as electrophysical characteristic signals.

[0029] Using electrophysical characteristic signals as excitation observations, and leveraging their transient response characteristics, a mapping relationship between current variation rate and voltage drop is established. The physical impedance parameters inside the energy storage are determined through online iterative convergence to identify dynamic equivalent parameters. Furthermore, multi-dimensional vector fusion is performed with real-time operating data to construct a global operating condition state space that reflects the energy exchange state and physical degradation trend.

[0030] It should be noted that the electrophysical characteristic signal is used as the excitation observation value. Based on Kirchhoff's voltage law, an analytical relationship is established that includes the energy storage terminal voltage, open circuit voltage, ohmic internal resistance, polarization resistance, and polarization capacitance. The transient response characteristics of the internal electrochemical process of energy storage at the microsecond time scale are captured by the electrophysical characteristic signal under high-frequency sampling. By calculating the current change per unit time (50 microseconds in the example), a mapping relationship between the current change rate and voltage drop is established, which reflects the linear combination relationship between the terminal voltage change and the current change rate, current amplitude, and the charging and discharging state of the polarization capacitor. By employing an iterative optimization strategy to iteratively converge the mapping relationship online, the unknown coefficients consisting of ohmic internal resistance, polarization resistance, and the product of polarization resistance and polarization capacitance are determined. Through extracting the converged unknown coefficients and performing algebraic decoupling operations, the values ​​of ohmic internal resistance, polarization resistance, and polarization capacitance within the energy storage are identified to determine the dynamic equivalent parameters. The identified dynamic equivalent parameters are then fused with real-time operating data through multi-dimensional spatial vector mapping to construct a global operating condition state space.

[0031] Methods for probabilistic interval prediction of source load power include: The low-frequency components are serialized and correlated with real-time operating data on the time axis, and dynamic equivalent parameters are fused to extract the spatiotemporal features of source-load evolution.

[0032] It should be noted that the timing clock generated by the high-precision crystal oscillator aligns and sorts the low-frequency components and real-time operating data on a unified time axis to form a source-load operating time series; the dynamic equivalent parameters are used as energy storage physical constraints, and after standardization and normalization, they are used together with the source-load operating time series to construct a multi-dimensional observation matrix. The multidimensional observation matrix is ​​segmented and the mean, variance, and skewness characteristics of power fluctuations within each sampling window are calculated. Principal component analysis is then used to reduce the dimensionality and decouple the multidimensional observation matrix, which contains the correlation between photovoltaic power fluctuations, load variation trends, and energy storage physical state degradation. Uncorrelated multidimensional feature vectors are extracted, and spatiotemporal features of source-load evolution that characterize the spatiotemporal evolution of energy exchange nodes are extracted.

[0033] Based on the historical evolution trajectory of the source-load evolution spatiotemporal characteristics, the expected trajectory of future power is extrapolated, and the prediction error sequence is determined by comparing with the historical measured power, thus obtaining the probability interval of the source-load power.

[0034] It should be noted that the spatiotemporal characteristics of source load evolution are subjected to multi-order difference processing to eliminate non-stationary trends; the differenced data are substituted into the Gaussian probability density function to construct a probability function, and within the parameter space determined by the order of the autoregressive term, a nonlinear search strategy is used to find the optimal solution that maximizes the joint probability density, and the autoregressive coefficient is obtained as an example of 0.85. The inertial trend is extracted by using a weighted linear combination of autoregressive coefficients to analyze historical observations. A weighted correction factor is then used to correct the historical prediction residuals. The expected power trajectory for a predetermined time period (15 minutes in the example) is extrapolated along the time axis through recursive iteration. Error samples are obtained by subtracting the historical measured power from the expected trajectory. 96 sampling points are used within a window to calculate the rolling variance and mean to determine the prediction error sequence. The preset confidence level is determined based on the standard deviation. The width of the confidence interval at 95% is used in the example. This width is then superimposed above and below the expected trajectory to obtain the probability interval of the source load power.

[0035] Within the prediction bandwidth formed by the probability interval, based on the statistical distribution characteristics of the prediction error sequence, multiple random power trajectories fluctuating around the expected future power trajectory are generated as typical power evolution sequences and arranged according to probability weights to obtain the source-load prediction distribution sequence.

[0036] It should be noted that within the prediction bandwidth formed by the probability interval, the cumulative distribution function of the estimated error sequence is divided equally (100 intervals in the example). Random sampling is performed in each interval and the result is mapped back to the power space using an inverse function transformation to generate multiple random power trajectories that fluctuate around the expected future power trajectory, serving as typical power evolution sequences. The occurrence frequency of each random power trajectory is calculated using a smoothing kernel function and arranged in descending order of probability weight to obtain the source-load prediction distribution sequence.

[0037] In traditional integrated photovoltaic-storage-charging stations, energy management typically relies on static power allocation based solely on predicted power at predetermined points. This approach fails to effectively address the uncertainties caused by drastic random fluctuations in photovoltaic power and load. Furthermore, existing multi-objective optimization methods often overlook the accelerated degradation of energy storage lifespan due to microscopic electrochemical processes induced by random fluctuations, making it difficult to balance economic benefits with battery physical health. Therefore, this invention extracts the distribution proportion of each random fluctuation path in the probability space and analyzes the internal physical stress of the energy storage in real time. The specific steps are as follows: Methods for identifying the expected power deviation due to source load randomness include: The typical power evolution sequence is synchronized with the expected future power trajectory, and the real-time offset vector of the typical power evolution sequence relative to the expected future power trajectory is extracted.

[0038] It should be noted that, using the timing clock generated by the high-precision crystal oscillator as a reference, the sampling timestamp of each random power trajectory in the typical power evolution sequence is matched one-to-one with the predicted time point corresponding to the expected future power trajectory, ensuring that the two are in phase synchronization on the same time axis. The instantaneous power value of each random power trajectory at each sampling time is subtracted from the expected power value of the expected future power trajectory at the corresponding time. The real-time offset vector of each typical power evolution sequence relative to the expected future power trajectory is calculated. The real-time offset vector includes the power deviation magnitude, the deviation change direction, and the polarity characteristics determined by the deviation change direction at each prediction step size.

[0039] The real-time offset vector is mapped and associated with the probability weights corresponding to typical power evolution sequences to determine the contribution of each typical power evolution sequence to power uncertainty.

[0040] Specifically, the real-time offset vector is mapped one-to-one with the probability weight. By establishing the product mapping relationship between the real-time offset vector and the probability weight, the distribution ratio of different fluctuation trajectories in the overall probability space is measured, thereby determining the contribution of each typical power evolution sequence to power uncertainty.

[0041] Based on the degree of contribution, the real-time offset vectors of each typical power evolution sequence at the same prediction time are weighted and integrated to identify the expected power deviation caused by the randomness of source and load.

[0042] Specifically, based on the degree of contribution, at the same prediction time, the real-time offset vectors corresponding to all typical power evolution sequences are weighted and summed with their corresponding probability weights. The weighted and integrated numerical value is used to represent the comprehensive statistical average of all possible random fluctuation paths at the current prediction time, thereby identifying the expected power deviation caused by the randomness of the source and load.

[0043] The output power allocation command methods include: Based on the typical power evolution sequences and their probability weights, and combined with the station's operating loss parameters, an economic operating cost function for operational efficiency is constructed.

[0044] It should be noted that the station's operating loss parameters are obtained by retrieving the conversion efficiency curves of each converter device in the integrated photovoltaic-storage-charging station. These operating loss parameters include, for example, a transformer no-load loss of 2kW, an inverter rated loss ratio of 1.5%, and line transmission resistance loss. The predicted power value corresponding to each typical power evolution sequence is multiplied with the probability weight corresponding to each typical power evolution sequence to calculate the expected energy exchange revenue. The expected energy exchange revenue is then subtracted from the energy loss cost corresponding to the station operation loss parameter, thereby constructing an economic operating cost function that represents the maximization of comprehensive revenue under different random fluctuation scenarios.

[0045] By utilizing the expected value of power deviation and dynamic equivalent parameters, the physical stress damage of random fluctuations of source load on the internal physical impedance of energy storage is quantified, and an energy storage loss cost function is constructed.

[0046] Specifically, by using the expected value of power deviation as the excitation variable, and combining the values ​​of ohmic internal resistance, polarization resistance and polarization capacitance in the dynamic equivalent parameters, the thermodynamic impact and electrochemical polarization pressure of the unexpected current generated by random fluctuations of source load on the internal physical impedance of energy storage are calculated. It should be noted that the thermodynamic shock is caused by the Joule heating element. The representation, expressed as: ; in, It is the expected value of power deviation. It is the open-circuit voltage value in the dynamic equivalent parameters. It is the ohmic internal resistance value in the dynamic equivalent parameters. This is the duration of the random fluctuations in the source load; the example value is 100 milliseconds. It should be noted that the electrochemical polarization pressure is determined by the polarization voltage offset. The characterization satisfies the following differential analytical relation: ; in, It is an infinitesimal sampling time increment. It is a tiny increment of the polarization voltage offset. It is the polarization voltage reference value in the dynamic equivalent parameters. It is the polarization capacitance value in the dynamic equivalent parameters. It is the polarization resistance value in the dynamic equivalent parameters; This allows for the quantification of physical stress damage to the internal physical impedance of energy storage caused by random fluctuations in source and load. Physical stress damage includes, for example, ohmic internal resistance temperature rise deformation, polarization capacitor voltage overshoot stress, and polarization concentration gradient damage caused by impeded electrolyte diffusion. A linear correlation ratio between physical stress damage and energy storage depreciation cost is established. For example, by using an empirical formula for cyclic aging attenuation, the unit transient polarization pressure is equivalent to the increase in energy storage capacity attenuation rate, multiplied by the replacement cost per unit capacity of energy storage, to construct an energy storage loss cost function that characterizes the degree of energy storage lifetime loss.

[0047] The economic operating cost function and the energy storage loss cost function are weighted and integrated to form a global objective function. Then, an optimization decision space is established by combining real-time operating data to find the optimal power flow scheme for each branch and output power allocation instructions.

[0048] Specifically, the economic operating cost function and the energy storage loss cost function are linearly summed using preset weighting coefficients to form a global objective function that maximizes operational benefits while minimizing physical damage. The preset weighting coefficients are set based on the Delphi method or hierarchical analysis method, determined according to the weighting ratio of electricity price volatility sensitivity and energy storage asset protection priority. For example, an example of a preset weighting coefficient of 0.7 for the economic weight and 0.3 for the lifetime loss weight, with a sum of 1. Using real-time operational data, three-phase voltage and current are retrieved as initial boundary conditions. Based on the physical connection structure of the photovoltaic-storage-charging integrated station, Kirchhoff's current law is used to establish power balance constraints for AC bus nodes, including photovoltaics, energy storage, and load. Combining the rated capacity and voltage amplitude of each converter equipment to determine the current withstand limit, as well as the chargeable and discharge power range of the energy storage battery and the grid demand limit, the upper and lower limits of the power exchange values ​​of each node are defined. By mapping the power balance constraints, current amplitude constraints, and power range constraints to a multidimensional Euclidean space, the convex set region that satisfies all physical operating conditions is determined, thereby establishing an optimization decision space. The optimization process seeks the extreme value of the global objective function within the decision space. A negative gradient search direction is constructed by calculating the partial derivatives of the global objective function with respect to the power exchange values ​​of each branch, and the power allocation weights of each node are iteratively updated along this direction. When the difference between the power exchange values ​​generated in two adjacent iterations is less than the example value... At that time, the power exchange values ​​between the photovoltaic inverter, energy storage converter and AC bus node that make the global objective function reach the optimal solution are locked, the optimal power flow scheme of each branch is determined and the power allocation command is output. This invention constructs a global objective function that includes probability weights, current variation rate, and voltage drop mapping, and performs online extremum optimization. It aims to improve the overall profitability of energy exchange nodes under random fluctuation scenarios and, to some extent, quantify and mitigate energy storage depreciation at the microscopic physical stress level, such as ohmic temperature rise deformation and polarization concentration gradient. This provides a feasible technical path for achieving refined and proactive protection against transient physical damage to energy storage assets while ensuring efficient and economical system operation.

[0049] The methods for obtaining adaptive control operators include: By using dynamic equivalent parameters and power allocation commands, the baseline control strength required to suppress power fluctuations is determined; the slope of the expected power deviation changes over time is analyzed to extract dynamic fluctuation characteristics that characterize the severity of random disturbances in the source and load.

[0050] Specifically, by retrieving the power allocation command determined by the optimal power flow scheme of each branch, and combining it with the ohmic internal resistance, polarization resistance, and polarization capacitance values ​​in the dynamic equivalent parameters, the compensation voltage amplitude required for the energy storage converter to overcome the internal physical impedance of the energy storage when executing the power allocation command is calculated. The expression is: ; in, The command current value is calculated by converting the power allocation command and the open-circuit voltage value; The dynamic polarization voltage drop generated by the command current value on the RC circuit formed by the polarization resistance value and the polarization capacitance value. By using a proportional-integral control algorithm, the calculated compensation voltage amplitude is linearly combined with the power distribution command to determine the reference control strength required to suppress power fluctuations under steady-state conditions. By using higher-order difference operators to perform continuous numerical operations on the power deviation expectation value sequence, the evolution trend of power fluctuation energy in the frequency and time domains is captured. The extracted dynamic fluctuation features include, for example, the amplitude of the second derivative of power deviation, the perturbation power spectral density, and the peak value of random perturbation energy, thereby quantifying the transient impact intensity caused by the randomness of the source load.

[0051] The dynamic fluctuation characteristics are compared with the baseline control strength. When the dynamic fluctuation characteristics exceed the preset safe fluctuation threshold, the correction deviation of the control gain is determined. The correction deviation is then added to the baseline control strength to output the adaptive control operator.

[0052] Specifically, the dynamic fluctuation characteristics are compared with a preset safe fluctuation threshold in real time. When the dynamic fluctuation characteristics exceed the preset safe fluctuation threshold (the preset safe fluctuation threshold can be set according to the absolute value of the second derivative of the power deviation; example value is given), the system will detect the fluctuation. This indicates that the current random disturbance of the source load exceeds the steady-state control range. The expression for the nonlinear mapping function between the dynamic fluctuation characteristics and the control gain is established as follows: ; in, It is the correction deviation of the control gain. The maximum corrected gain limit is 0.25 as an example. This is the sensitivity adjustment coefficient. By collecting historical operating data and extracting typical source-load random disturbance samples, different sensitivity adjustment coefficients are traversed within the parameter space to lock onto the value that minimizes the sum of squares of the power allocation command tracking error. This value is then determined as the sensitivity adjustment coefficient, with an example value of 0.1. It is a value with dynamic fluctuation characteristics. It is a preset safety fluctuation threshold. It is a hyperbolic tangent function used to nonlinearly map the difference between dynamic fluctuation characteristics and safety threshold to a continuous interval of (-1, 1); The correction deviation of the control gain is directly superimposed on the reference control strength. By changing the proportional gain or feedforward compensation coefficient of the control loop, the execution step size and response speed of the power distribution command are adjusted to compensate for the transient impact of the expected power deviation caused by the randomness of the source load on the stability of the AC bus voltage. The output is an adaptive control operator with real-time disturbance resistance capability.

[0053] Traditional converter control typically employs a fixed proportional-integral gain and a constant filter cutoff frequency. When faced with uncertainties arising from photovoltaic and load fluctuations, this fixed-parameter control loop cannot simultaneously ensure rapid dynamic response and system stability, easily leading to electromagnetic disturbances or bus voltage drops. Therefore, this invention dynamically maps an adaptive control operator to the gain adjustment coefficient of the underlying loop, achieving dynamic adaptive scaling of the closed-loop bandwidth and duty cycle adjustment response time. The specific steps are as follows: The methods for dynamically adjusting the control gain based on adaptive control operators include: Using the power distribution command as a reference trajectory, vector alignment is performed with real-time operating data on a time scale to extract the initial error state of the converter's current regulation performance and the degree of environmental disturbance.

[0054] Specifically, the timing clock generated by the high-precision crystal oscillator is retrieved, and the expected power values ​​of each node preset in the power distribution command are used as reference trajectories to align with the real-time measured values ​​of three-phase voltage and three-phase current in the real-time operation data on a unified time axis to ensure spatial vector phase synchronization. By calculating the difference between the power allocation command and the actual power feedback value in the real-time operating data, the deviation characteristics of the current regulation performance of the converter in the dynamic response process are extracted. Combined with electrophysical characteristic signals, the instantaneous disturbance components generated by environmental disturbances on the energy exchange nodes are captured. The deviation characteristics and instantaneous disturbance components are then combined into a multi-dimensional vector synthesis to extract the initial error state of the current regulation performance of the converter and the degree of environmental disturbance.

[0055] The initial error state is input into the adaptive control operator and compared with the hardware control boundary to determine the gain adjustment coefficient for adjusting the sensitivity of the underlying loop.

[0056] Specifically, the initial error state is input into the adaptive control operator, and the initial error state is proportionally amplified by the reference control strength in the adaptive control operator, and the nonlinear compensation increment generated by the correction deviation of the control gain is superimposed. The maximum regulation duty cycle (example value is 0.95), rated current limit and inductor saturation current in the hardware manual of the converter equipment are retrieved as hardware control boundaries. The weighted and mapped signal is limited to the physical tolerance range of the hardware through amplitude clamping operation. If the calculated rate of change of current exceeds the inductor saturation limit, the output value is forcibly locked to the edge value of the hardware control boundary. By calculating the algebraic difference before and after signal mapping and dividing it by the discrete sampling period, the rate of change of the initial error state under the action of the adaptive control operator is obtained. This rate of change is then compared with the sensitivity reference value to determine the gain adjustment coefficient for adjusting the sensitivity of the underlying loop.

[0057] By using the gain adjustment coefficient to update the control parameters of the converter's internal loop in real time, the control gain can be dynamically adjusted according to the adaptive control operator.

[0058] Specifically, the control parameters of the converter's internal loop are updated in real time using the gain adjustment coefficient. The gain adjustment coefficient is directly multiplied with the proportional gain and integral gain of the converter's internal loop. In each pulse width modulation cycle, the comparison register value of the converter's internal pulse width modulation generator is recalculated based on the updated gain, thereby changing the duty cycle adjustment step size. By shortening or lengthening the duty cycle adjustment response time, the tracking sensitivity of the energy storage converter and photovoltaic inverter to power distribution commands can be changed. For example, under the condition of pulse width modulation frequency of 10kHz (i.e., period of 100 microseconds), the duty cycle adjustment response time refers to the total discrete length required for the control loop to update the duty cycle register from detecting power deviation. When the dynamic fluctuation characteristics are severe, the duty cycle adjustment response time can be shortened from 500 microseconds (i.e., 5 modulation cycles) in steady state to 200 microseconds (i.e., 2 modulation cycles) by increasing the gain adjustment coefficient, so as to achieve high-speed transient response. When the source load fluctuation is small, the duty cycle adjustment response time can be lengthened from 500 microseconds to 1.5 milliseconds (i.e., 15 modulation cycles) by decreasing the gain adjustment coefficient, so as to increase loop damping and suppress unnecessary electromagnetic disturbances. The gain adjustment coefficient directly changes the transfer function gain of the underlying current control loop, enabling the closed-loop bandwidth of the converter to dynamically expand and contract with the random fluctuations of the source load. This compensates for the impact of the expected power deviation caused by the randomness of the source load on the stability of the AC bus voltage, ensuring that the dynamic response characteristics of the converter remain consistent with the disturbance rejection capability preset by the adaptive control operator, and realizing the dynamic adjustment of the control gain according to the adaptive control operator.

[0059] Methods for generating converter drive signals include: The control loop bandwidth of the converter is adjusted by using the dynamically adjusted control gain, and the power distribution command is smoothed by filtering to obtain a smooth power reference.

[0060] Specifically, by calculating the scaling effect of the dynamically adjusted control gain on the time constant of the low-pass filter, the gain adjustment coefficient is applied to the transfer function of the converter's internal loop. This allows the converter's control loop bandwidth to smoothly switch between a preset low-frequency band (e.g., 10Hz in this example) and a high-frequency band (e.g., 500Hz in this example), enabling the converter's closed-loop cutoff frequency to dynamically shift with the random fluctuations of the source and load energy levels. The adjusted control loop bandwidth is then used to smooth the power distribution command, filtering out transient power jumps that exceed the current converter's response capability, thus obtaining a smooth power reference that conforms to the converter's dynamic execution capability.

[0061] By comparing real-time operating data with a smoothed power reference, a compensation current vector is generated to offset transient current surges. This vector is then superimposed on the target control voltage converted from the smoothed power reference to form a comprehensive control voltage command. Specifically, the three-phase current measurement values ​​in the real-time operating data are subtracted from the reference current values ​​in real time to extract the transient current deviation caused by the randomness of the source load; the transient current deviation is quickly calculated using proportional-derivative control logic to generate a compensation current vector with reverse cancellation effect, and the compensation current vector is mapped into a voltage compensation term according to Kirchhoff's voltage law; this voltage compensation term is directly superimposed on the target control voltage obtained by smoothing the power reference, and by canceling the dynamic voltage drop caused by the internal physical impedance of the energy storage in real time, the output voltage command includes pre-compensation for the electrochemical polarization and ohmic loss inside the energy storage, thereby forming a comprehensive control voltage command with disturbance self-suppression characteristics.

[0062] The comprehensive control voltage command is compared with the preset high-frequency carrier signal in real time, and the polarity of the output level is triggered to generate the converter drive signal based on the amplitude relationship between the comprehensive control voltage command and the preset high-frequency carrier signal.

[0063] Specifically, a preset high-frequency carrier signal (in this example, a triangular carrier signal with a value of 10kHz) is retrieved, and within each pulse width modulation cycle, the instantaneous amplitude of the comprehensive control voltage command is used as a comparison reference value and compared with the real-time slope value of the triangular carrier signal in real time. Based on the amplitude relationship between the integrated control voltage command and the preset high-frequency carrier signal, the logic level of the voltage comparator switches. When the amplitude of the integrated control voltage command is higher than that of the triangular carrier signal, the internal power switch driver circuit of the converter is triggered, and the output level flips upward. When the amplitude of the integrated control voltage command is lower than that of the triangular carrier signal, the output level flips downward. By precisely controlling the timing of the polarity flip within a single modulation cycle, the duty cycle of the output level is adjusted, ultimately generating the converter drive signal used to drive the switching action of the power semiconductor devices. This invention constructs a dynamic gain adjustment and pulse width modulation generation mechanism that includes an initial error state, hardware control boundaries, and a high-frequency carrier signal. This not only improves the converter's tracking sensitivity to power allocation commands on a microsecond-level time scale, but also alleviates the transient impact of random source-load fluctuations on the stability of the AC bus voltage to a certain extent. This provides a feasible technical path to enhance the disturbance resistance capability of the integrated photovoltaic-storage-charging station under complex random operating conditions while ensuring the physical safety of the hardware.

[0064] Methods for calculating deviation residuals include: Monitor the feedback state after the converter drive signal is applied; and perform multi-dimensional vector alignment with the expected response state composed of power allocation command and future power expectation trajectory to determine the multi-dimensional error vector.

[0065] Specifically, the process involves: real-time monitoring of the feedback state after the converter drive signal is applied using electrical sensors to obtain real-time measurements of the three-phase voltage and three-phase current of the energy storage converter and photovoltaic inverter after the execution of the drive command; retrieving the timing clock generated by a high-precision crystal oscillator to align the real-time measurements with the expected response state, which is composed of the power distribution command and the expected future power trajectory, on a unified time axis to ensure that the sampling times of the feedback state and the expected response state are completely synchronized; calculating the algebraic difference between the feedback state and the expected response state in the dimensions of power amplitude, voltage phase, and current change rate to construct a multi-dimensional error vector characterizing the deviation of the current controlled object from the expected operating trajectory, thereby determining the multi-dimensional error vector.

[0066] By using dynamic equivalent parameters, the steady-state component caused by inherent energy storage losses is separated from the initial error vector, and the dynamic deviation characteristics caused by random disturbances of source and load are extracted to obtain the deviation residual.

[0067] Specifically, by using the values ​​of ohmic internal resistance, polarization resistance, and polarization capacitance in the dynamic equivalent parameters, the inherent voltage drop and polarization loss generated by the physical impedance inside the energy storage under the current excitation are calculated; the steady-state response consisting of the inherent voltage drop and polarization loss is subtracted from the multidimensional error vector, thereby separating the steady-state component generated by the inherent loss of energy storage. By extracting high-frequency features from the remaining error signals, the nonlinear deviation caused by transient impacts of power fluctuations at the source and load ends is identified and defined as a dynamic deviation feature caused by random disturbances of the source and load. The Euclidean norm of the dynamic deviation feature is calculated by multidimensional spatial mapping to obtain the deviation residual characterizing the real-time runaway energy level of the system.

[0068] The parameters relating the global operating condition state space to the optimization objective are corrected as follows: The deviation residual is mapped to the global operating condition state space to determine the correction vector; and then superimposed onto the global operating condition state space to update the weights of the electrophysical characteristic signals in the global operating condition state space.

[0069] It should be explained that the deviation residual is mapped to the global operating condition state space. Using the deviation residual as a feedback adjustment, a correction vector is calculated to correct the coupling relationship between the physical degradation trend and the energy exchange state. This correction vector is then applied algebraically to the multidimensional observation matrix of the global operating condition state space. By adjusting the scaling factors of the electrophysical characteristic signals in the global operating condition state space mapping matrix, the weights of the electrophysical characteristic signals in the global operating condition state space are updated. This ensures that the global operating condition state space can reflect the dynamic evolution of physical impedance caused by the randomness of the source and load in real time. Based on the deviation residuals, the weight allocation of the global objective function is adjusted according to the proportion of economic operating costs and energy storage losses to determine the sensitivity coefficient; combined with the calibrated global operating state space, the input benchmark and target constraints of the optimization action are updated synchronously to achieve dynamic correction of the associated parameters.

[0070] It should be noted that, based on the magnitude of the deviation residual, the distribution ratio of the deviation residual in economic operating costs and energy storage loss costs is calculated in real time. When the deviation residual is mainly caused by power fluctuations, the priority of energy storage protection is increased by increasing the preset weighting coefficient of the lifetime loss weight; when the deviation residual is within the preset safety fluctuation threshold, the priority of economic benefits is increased by increasing the preset weighting coefficient of the economic weight, thereby adjusting the weight allocation of the global objective function to determine the sensitivity coefficient. Combining the global operating state space after weight update, the calibrated dynamic equivalent parameters are simultaneously used as the input benchmark for the optimization action, and the updated power balance constraints and physical stress damage limits are used as target constraints. The search boundary of the optimization decision space is redefined using the negative gradient search direction to achieve dynamic correction of the associated parameters.

[0071] In summary, this invention achieves a logical closed loop of "deep physical state perception" and "precise stress loss offsetting" by extracting high-frequency electrophysical characteristic signals to identify dynamic equivalent parameters of energy storage online and constructing a global operating condition state space. It also incorporates the quantification of stress damage caused by random fluctuations to physical impedance to construct an energy storage loss cost function. This transforms the fuzzy randomness of source and load into quantifiable physical stress indicators and utilizes adaptive control operators to drive the converter for flexible smoothing and transient current compensation. This addresses the conflict between operational efficiency and equipment lifespan in integrated photovoltaic-energy storage-charging stations under extreme operating conditions, enabling the control system to possess "physical self-sensing" and "operational self-evolution" capabilities. While minimizing the total lifespan operating cost, it ensures the physical safety and service life of energy storage in complex random environments.

[0072] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty, characterized in that, include: The state construction module acquires real-time operating data of the integrated photovoltaic-storage-charging station, extracts electrophysical characteristic signals during the power conversion process to identify the dynamic equivalent parameters of energy storage online, and constructs a global operating condition state space. The prediction module performs probability interval prediction of source load power based on the global operating condition state space, obtains the source load prediction distribution sequence, and identifies the expected value of power deviation caused by the randomness of source load. The optimization module combines the source-load predicted distribution sequence with the expected power deviation value to optimize the economic operating cost and energy storage loss, and outputs power allocation instructions and adaptive control operators. The execution module dynamically adjusts the control gain according to the adaptive control operator to smooth the power distribution command and reduce the transient physical stress of uncertainty fluctuations on energy storage, and generates the converter drive signal. The self-evolution module monitors the feedback state after the converter drive signal is applied and calculates the deviation residual, dynamically correcting the correlation parameters between the global operating condition state space and the optimization objective.

2. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 1, characterized in that, The methods for constructing the global operating condition state space include: The electrical parameters of the energy exchange nodes in the integrated photovoltaic, energy storage and charging station are collected to obtain real-time operating data, and the data is aligned with the ambient temperature data on a time scale. Modal decomposition is performed on real-time operating data to separate the high-frequency components generated by switching actions and the low-frequency components of power fluctuations, and the high-frequency components are identified as electrophysical characteristic signals. Using electrophysical characteristic signals as excitation observations, and leveraging their transient response characteristics, a mapping relationship between current variation rate and voltage drop is established. The physical impedance parameters inside the energy storage are determined through online iterative convergence to identify dynamic equivalent parameters. Furthermore, multi-dimensional vector fusion is performed with real-time operating data to construct a global operating condition state space that reflects the energy exchange state and physical degradation trend.

3. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 2, characterized in that, The methods for probabilistic interval prediction of source load power include: The low-frequency components are serialized and correlated with real-time operating data on the time axis, and dynamic equivalent parameters are fused to extract the spatiotemporal features of source load evolution. Based on the historical evolution trajectory of the source-load evolution spatiotemporal characteristics, the expected trajectory of future power is extrapolated, and the prediction error sequence is determined by comparing with the historical measured power, thus obtaining the probability interval of the source-load power. Within the prediction bandwidth formed by the probability interval, based on the statistical distribution characteristics of the prediction error sequence, multiple random power trajectories fluctuating around the expected future power trajectory are generated as typical power evolution sequences and arranged according to probability weights to obtain the source-load prediction distribution sequence.

4. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 3, characterized in that, The methods for identifying the expected power deviation due to source load randomness include: The typical power evolution sequence is synchronized with the expected future power trajectory, and the real-time offset vector of the typical power evolution sequence relative to the expected future power trajectory is extracted. The real-time offset vector is mapped and associated with the probability weights corresponding to typical power evolution sequences to determine the contribution of each typical power evolution sequence to power uncertainty. Based on the degree of contribution, the real-time offset vectors of each typical power evolution sequence at the same prediction time are weighted and integrated to identify the expected power deviation caused by the randomness of source and load.

5. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 4, characterized in that, The output power allocation command includes the following methods: Based on each typical power evolution sequence and its probability weight, and combined with the station's operating loss parameters, an economic operating cost function for operational efficiency is constructed. By utilizing the expected value of power deviation and dynamic equivalent parameters, the physical stress damage of random source load fluctuations on the internal physical impedance of energy storage is quantified, and an energy storage loss cost function is constructed. The economic operating cost function and the energy storage loss cost function are weighted and integrated to form a global objective function. Then, an optimization decision space is established by combining real-time operating data to find the optimal power flow scheme for each branch and output power allocation instructions.

6. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 5, characterized in that, The methods for obtaining the adaptive control operator include: By using dynamic equivalent parameters and power allocation commands, the baseline control strength required to suppress power fluctuations is determined; the slope of the expected power deviation changes over time is analyzed to extract dynamic fluctuation characteristics that characterize the severity of random disturbances in the source load. The dynamic fluctuation characteristics are compared with the baseline control strength. When the dynamic fluctuation characteristics exceed the preset safe fluctuation threshold, the correction deviation of the control gain is determined. The correction deviation is then added to the baseline control strength to output the adaptive control operator.

7. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 6, characterized in that, The method of dynamically adjusting the control gain according to the adaptive control operator includes: Using the power distribution command as a reference trajectory, the vector alignment with the real-time operating data is performed on a time scale to extract the initial error state of the converter's current regulation performance and the degree of environmental disturbance. The initial error state is input into the adaptive control operator and compared with the hardware control boundary to determine the gain adjustment coefficient for adjusting the sensitivity of the underlying loop. By using the gain adjustment coefficient to update the control parameters of the converter's internal loop in real time, the control gain can be dynamically adjusted according to the adaptive control operator.

8. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 7, characterized in that, The methods for generating converter drive signals include: The control loop bandwidth of the converter is adjusted by using the dynamically adjusted control gain, and the power distribution command is smoothed by filtering to obtain a smooth power reference. By comparing real-time operating data with a smoothed power reference, a compensation current vector is generated to offset transient current surges. This vector is then superimposed on the target control voltage converted from the smoothed power reference to form a comprehensive control voltage command. The comprehensive control voltage command is compared with the preset high-frequency carrier signal in real time, and the polarity of the output level is triggered to generate the converter drive signal based on the amplitude relationship between the comprehensive control voltage command and the preset high-frequency carrier signal.

9. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 8, characterized in that, The methods for calculating the deviation residuals include: Monitor the feedback state after the converter drive signal is applied; and perform multi-dimensional vector alignment with the expected response state composed of power allocation command and future power expectation trajectory to determine the multi-dimensional error vector; By using dynamic equivalent parameters, the steady-state component caused by inherent energy storage losses is separated from the initial error vector, and the dynamic deviation characteristics caused by random disturbances of source and load are extracted to obtain the deviation residual.

10. The adaptive control system for an integrated photovoltaic-storage-charging station considering source-load uncertainty as described in claim 9, characterized in that, The parameters relating the corrected global operating condition state space to the optimization objective include: The deviation residual is mapped to the global operating condition state space to determine the correction vector; and then superimposed onto the global operating condition state space to update the weights of the electrophysical characteristic signals in the global operating condition state space. Based on the deviation residuals, the weight allocation of the global objective function is adjusted according to the proportion of economic operating costs and energy storage losses to determine the sensitivity coefficient; combined with the calibrated global operating state space, the input benchmark and target constraints of the optimization action are updated synchronously to achieve dynamic correction of the associated parameters.